Download Schumpeter Meeting Keynes: A Policy-Friendly Model of

Working Paper Series
Department of Economics
University of Verona
Schumpeter Meeting Keynes: A Policy-Friendly Model of
Endogenous Growth and Business Cycles
Giovanni Dosi, Giorgio Fagiolo, Andrea Roventini
WP Number: 50
ISSN:
October 2008
2036-2919 (paper),
2036-4679 (online)
Schumpeter Meeting Keynes: A Policy-Friendly Model
of Endogenous Growth and Business Cycles∗
Giovanni Dosi†
Giorgio Fagiolo‡
Andrea Roventini§
October 22, 2008
Abstract
This paper studies an agent-based model that bridges Keynesian theories of demandgeneration and Schumpeterian theories of technology-fueled economic growth. We
employ the model to investigate the properties of macroeconomic dynamics and
the impact of public polices on supply, demand and the “fundamentals” of the
economy. We find that the complementarities between factors influencing aggregate
demand and drivers of technological change affect both “short-run” fluctuations
and long-term growth patterns. From a normative point of view, simulations show
a corresponding complementarity between Keynesian and Schumpeterian policies in
sustaining long-run growth paths characterized by mild fluctuations and acceptable
unemployment levels. The matching or mismatching between innovative exploration
of new technologies and the conditions of demand generation appear to suggest the
presence of two distinct “regimes” of growth (or absence thereof) characterized by
different short-run fluctuations and unemployment levels.
Keywords: Endogenous Growth; Business Cycles; Growth Policies; Business Cycle
Policies; Evolutionary Economics; Agent-Based Computational Economics; PostWalrasian Economics; Empirical Validation; Monte-Carlo Simulations.
JEL Classification: E32, E6, O3, O4
∗
Thanks to Mario Cimoli, Herbert Dawid, Andrea Ginzburg, Paola Giuri, David Lane, Stan Metcalfe,
Andreas Pyka, Pier Paolo Saviotti, and Bart Verspagen. Thanks also to the participants to the 12th
International Joseph A. Schumpeter Society conference, Rio de Janeiro, Brazil; the 14th International
Conference on Computing in Economics and Finance, Paris, France; and the “Economics of Innovative
Change” summer school organized by the Max Planck Institute and the Friedrich Schiller University,
Jena, Germany. Previous versions of the paper have also benefited from comments received during the
5th International Globelics conference, Saratov, Russia; ‘the ‘Economics of Technology Policy” conference
at Monte Verità, Ascona, Switzerland; the 5th International EMAEE conference, Manchester, U.K.; the
Brisbane Club meeting held during the 4th Storep conference, Pollenzo, Italy; and the “Institutional and
Social Dynamics of Growth and Distribution” conference, Lucca, Italy. All usual disclaimers apply.
†
Corresponding Author. Sant’Anna School of Advanced Studies, Pisa, Italy and MIoIR, University of
Manchester, U.K. Mail address: Sant’Anna School of Advanced Studies, Piazza Martiri della Libertà 33,
I-56127 Pisa, Italy. Tel: +39-050-883326. Fax: +39-050-883344. Email: [email protected]
‡
Sant’Anna School of Advanced Studies, Pisa, Italy. Mail address: Sant’Anna School of Advanced
Studies, Piazza Martiri della Libertà 33, I-56127 Pisa, Italy. Tel: +39-050-883282. Fax: +39-050-883344.
Email: [email protected]
§
University of Verona, Italy and Sant’Anna School of Advanced Studies, Pisa, Italy. Mail address:
Università di Verona, Dipartimento di Scienze Economiche, viale dell’Università 4, I-37129 Verona, Italy.
Tel: +39-045-8028238. Fax: +39-045-8028529. Email: [email protected]
1
1
Introduction
This work studies an agent-based model (ABM) of endogenous growth and business cycles
and explores its properties under different public policies impacting on supply, demand,
and the “fundamentals” of the economy. We extend the model presented in Dosi et al.
(2006, 2008), which we use also as a sort of “policy laboratory” where both business-cycle
and growth effects of alternative public interventions may be evaluated under different
techno-economic scenarios.
A fundamental feature of the model is that it bridges Keynesian theories of demand
fluctuations and Schumpeterian theories of economic growth. The model also allows to
experiment with an ensemble of policies, related to the structural features of the economy
(concerning e.g., technology, industry structure and competition) on the one hand and to
demand macro-management, on the other.
Historically, a major divide has emerged in macroeconomics theories. Long-run approaches have traditionally dealt with growth issues in a strict sense, trying to account
for (broken-linear or stochastic) trends present in macro time series, while leaving to
“short-run” models the task of explaining economic fluctuations around the trend. An
early example is the way the IS-LM interpretation of Keynes (Hicks, 1937) and the models rooted in Solow (1956) found their division of labor addressing business cycles and
growth, respectively1 .
Since then, the balance has been shifting over time. At one extreme, the “new classical
economics” has boldly claimed the irrelevance of any “Keynesian” feature of the economy.
New Keynesian models have defended the turf of “non-fundamental” fluctuations most
often on the grounds of informational and behavioural frictions (an insightful overview is
in Blanchard, 2008), with just a minority holding the view that such “imperfections” are
in fact structural, long-term characteristics of the economy (see Akerlof and Yellen, 1985;
Greenwald and Stiglitz, 1993a,b; Akerlof, 2002, 2007, among them). Lacking a better
name, let we call the latter Hard New Keynesians, HNK henceforth.
More recently, the new neoclassical synthesis between real business cycle (RBC) and a
major breed of New Keynesian models has refined the interactions and the territorial divisions between “fundamental dynamics” and higher frequency, “non-fundamental” shocks
within the Dynamic Stochastic General-Equilibrium (DSGE) theoretical family (cf. the
classic Woodford, 2003; Galı́ and Gertler, 2007). In fact, DSGE models feature a core
with an RBC engine to which one may easily add sticky prices, imperfect competition,
monetary policy (Taylor-like) rules, and whatever can be imaginatively squeezed into
the underlying “structural model”2 . Indeed, there is hardly any Schumpeter in term of
1
For an interesting reconstruction of the econometric counterpart of such a divide in the 30’s and 40’s
debate, see Louca (2001).
2
As Blanchard (2008) puts it, “To caricature only slightly: a macroeconomic article today follows
strict, haiku-like, rules: it starts from a general equilibrium structure, in which individuals maximize
2
endogenous innovation in DSGE models.
From a quite different angle, endogenous growth models, notwithstanding very different features (from Romer, 1990 to Aghion and Howitt, 1992 and Dinopoulos and
Segerstrom, 1999), possess an implicit or explicit Schumpeterian engine: innovation and
thus the dynamics in the “technological fundamentals” of the economy is endogenous.
At the same time, “non-fundamental” (e.g. demand-related) fluctuations do not appear
in this family of models. Refinements, such as Aghion and Howitt (1998)3 , do entail
equilibrium fluctuations wherein Keynesian features do not play any role4 .
Somewhat similarly, evolutionary models, as pioneered by Nelson and Winter (1982),
are driven by a Schumpeterian core with endogenous innovation, but do largely neglect
too any demand-related driver of macroeconomic activity5 .
The model which follows, shares evolutionary roots, but in tune with HNK insights (cf.
for example Stiglitz, 1993) tries to explore the feedbacks between the factors influencing
aggregate demand and those driving technological change. By doing that we begin to offer
a unified framework jointly accounting for long-term dynamics and higher frequencies
fluctuations.
The model is certainly post-Walrasian (Colander, 2006; Colander et al., 2008) meaning
that it goes beyond the purported Walrasian foundations squeezed into the representative agent assumption nested in DSGE models and the general commitment to market
clearing. In that, well in tune with the growing literature on agent-based computational
economics (ACE; see Tesfatsion and Judd, 2006; LeBaron and Tesfatsion, 2008), the
model meets Solow’s plea for micro heterogeneity (Solow, 2008): a multiplicity of agents
interact without any ex ante commitment to the reciprocal consistency of their actions6 .
Furthermore, the model — alike most evolutionary ABMs — is “structural” in the
sense that it explicitly builds on a representation of what agents do, how they adjust, etc.
In that, it is as far as the DSGE perspective from “old Keynesian” models studying the
relations amongst aggregate variables without any explicit microfoundation. At the same
time, our commitment is to “phenomenologically” describe microbehaviours as close as
one can get to available micro evidence. Akerlof’s advocacy of a “behavioural microeconomics”, we believe, builds on that notion (Akerlof, 2002). In fact, this is our first
fundamental disciplining device. A second, complementary one involves the ability of the
model to jointly account for an ensemble of stylized facts regarding both “micro/meso”
the expected present value of utility, firms maximize their value, and markets clear. Then, it introduces
a twist, be it an imperfection or the closing of a particular set of markets, and works out the general
equilibrium implications. It then performs a numerical simulation, based on calibration, showing that
the model performs well. It ends with a welfare assessment.” (p. 26).
3
See also Aghion et al. (2005); Aghion and Marinescu (2007); Aghion et al. (2008).
4
Ironically, given the lack of stability of “new growth” trajectories, “Keynesianism” could show its
full force: see also below.
5
See however Dosi et al. (1994) and Fagiolo and Dosi (2003) for exceptions.
6
For germane ABMs with both some Keynesian and Schumpeterian elements see Verspagen (2002),
Ciarli et al. (2008), Saviotti and Pyka (2008), and the discussion in Silverberg and Verspagen (2005).
3
aggregates such as indicators of industrial structures (e.g. firm size distributions, productivity dispersions, firm growth rates) together with macro statistical properties (including
rates of output growth, output volatility, unemployment rates, etc.).
Our work shares many ingredients with (and in many ways is complementary to) the research project carried on within the European project EURACE (http://www.eurace.org),
which features a large-scale ABM aiming at capturing the main characteristics of the
European economy and addressing European-policy analyses (Deissenberg et al., 2008;
Dawid et al., 2008). Unlike EURACE models, however, we keep the scale of the system
relatively small, in line with traditional macroeconomic ABMs with little overall calibration exercises, albeit with attention to empirically sound micro rules and interaction
mechanisms.
The model below describes an economy composed of firms, consumers/workers and
a public sector. Firms belong to two industries. In the first one, firms perform R&D
and produce heterogeneous machine tools. Firms in the second industry invest in new
machines and produce a homogenous consumption good. Consumers sell their labor to
firms in both sectors and fully consume the income they receive. The government levies
taxes on workers’ wages and firms’ profits and it provides unemployed workers with a
fraction of the market wage.
As customary in evolutionary/ACE perspectives, the policy framework studied here
is explored via computer simulations. To overcome the well-known problems related to
sensitivity to the choice of parameters, possibly arising in ABMs7 , we look for policy implications that: (i) are robust to reasonable changes in the parameters of the model; (ii)
refer to model setups and parametrizations wherein the output of the model is empirically
validated (i.e., simulated microeconomic and macroeconomic data possess statistical properties similar to those empirically observed in reality). We consider this as a value added
of our study, as very often in the literature policy experiments are performed without
imposing any ex-ante empirical-validation requirement on the model (Fukac and Pagan,
2006; Canova, 2008; Fagiolo and Roventini, 2008). Policy configurations are captured by
different “control” parameters and different institutional, market or industry setups. The
impact of different policies is then quantitatively assessed in terms of ensuing aggregates
such as average output growth, output volatility, average unemployment, etc.
One of the main insights stemming from our extensive policy-simulation exercises is
a vindication of a strong complementarity between Schumpeterian policies addressing innovative activities and Keynesian demand management policies. Both types of policies
seem to be necessary to put the economy into a long-run sustained growth path. Schumpeterian policies potentially foster economic growth, but they do not appear to be able
7
See Fagiolo et al. (2007) for a discussion; more on that in Section 3. The potential for policy exercises
in ABMs is discussed in the special issue on “Agent-Based Models for Economic Policy Design” of the
Journal of Economic Behavior and Organization, 2008 (vol. 67, no. 2), edited by Herbert Dawid and
Giorgio Fagiolo.
4
alone to yield sustained long-run growth. In a broad parameter region, “fundamental”
(indeed, endogenously generated) changes in technology are unable to fully propagate in
terms of demand generation and ultimately output growth. By the same token, demand
shocks (in the simplest case, induced by government fiscal policies) bear persistent effects
upon output levels, rates of growth, and rates of innovations. In that, Keynesian policies
not only have a strong impact on output volatility and unemployment but seem to be
also a necessary condition for long-run economic growth.
In fact, our results suggest that the matching or mismatching between innovative exploration of new technologies and the conditions of demand generation appear to yield two
distinct “regimes” of growth (or absence thereof), also characterized by different short-run
fluctuations and unemployment levels. Even when Keynesian policies allow for a sustained
growth, their tuning affects the amplitude of fluctuations and the long-term levels of unemployment and output. Symmetrically, fluctuations and unemployment rates are also
affected by “Schumpeterian policies”, holding constant macro demand management rules.
The rest of the paper is organized as follows. Section 2 describes the model. In Section
3 we perform empirical validation checks and in Section 4 we present results on policy
exercises. Finally, Section 5 concludes and discusses future extensions.
2
The Model
As already mentioned, our simple economy is composed of a machine-producing sector
made of F1 firms (denoted by the subscript i), a consumption-good sector made of F2
firms (denoted by the subscript j), LS consumers/workers, and a public sector. Capitalgood firms invest in R&D and produce heterogenous machines. Consumption-good firms
combine machine tools bought by capital-good firms and labor in order to produce a
final product for consumers. The public sector levies taxes on firms’ profits and pay
unemployment benefits. Innovations is clearly endogenous to our economy. It is the
uncertain outcome of the search efforts of the producers of capital equipment and exerts
its impact throughout the economy via both the lowering of the production costs of
such equipment and its diffusion in the “downstream” consumption-good sector. Before
accurately describing the model, we briefly provide the timeline of events occurring in
each time step.
2.1
The Timeline of Events
In any given time period (t), the following microeconomic decisions take place in sequential
order:
1. Machine-tool firms perform R&D trying to discover new products and more efficient production techniques and to imitate the technology and the products of their
5
competitors.
2. Capital-good firms advertise their machines with consumption-good producers.
3. Consumption-good firms decide how much to produce and invest. If investment is
positive, consumption-good firms choose their supplier and send their orders.
4. In both industries firms hire workers according to their production plans and start
producing.
5. Imperfectly competitive consumption-good market opens. The market shares of
firms evolve according to their price competitiveness.
6. Entry and exit take places. In both sectors firms with near zero market shares and
negative net liquid assets are eschewed from the two industries and replaced by new
firms.
7. Machines ordered at the beginning of the period are delivered and become part of
the capital stock at time t + 1.
At the end of each time step, aggregate variables (e.g. GDP, investment, employment)
are computed, summing over the corresponding microeconomic variables.
Let us now turn to a more detailed description of the model and of the agents’ behaviours, which — to repeat — we try to keep as close as we can to what we know they
actually do as distinct from what they ought to do under more perfect informational
circumstances.
2.2
The Capital-Good Industry
The technology of a capital-good firms is (Aτi , Biτ ), where the former coefficient stands for
the labor productivity of the machine-tool manufactured by i for the consumption-good
industry (a rough measure of producer quality), while the latter coefficient is the labor
productivity of the production technique employed by firm i itself. The positive integer
τ denotes the current technology vintage. Given the monetary wage w, the unit cost of
production of capital-good firms is:
ci (t) =
w(t)
.
Biτ
(1)
With a fixed mark-up (µ1 > 0) pricing rule8 , prices (pi ) are defined as:
pi (t) = (1 + µ1 )ci (t).
8
(2)
Survey data evidence summarized in Fabiani et al. (2006) show that European firms mostly set
prices according to mark-up rules .
6
The unit labor cost of production in the consumption-good sector associated with each
machine of vintage τ , produced by firm i is:
c(Aτi , t) =
w(t)
.
Aτi
Firms in the capital-good industry “adaptively” strive to increase their market shares
and their profits trying to improve their technology both via innovation and imitation.
Both are costly processes: firms invest in R&D a fraction of their past sales (Si ):
RDi (t) = νSi (t − 1),
(3)
with 0 < ν < 1. R&D expenditures are employed to hire researchers paying the market
wage w(t)9 . Firms split their R&D efforts between innovation (IN ) and imitation (IM )
according to the parameter ξ ∈ [0, 1]10 :
INi (t) = ξRDi (t)
IMi (t) = (1 − ξ)RDi (t).
We model innovation as a two steps process. The first one determines whether a
firm obtains or not an access to innovation — irrespectively of whether it is ultimately
a success or a failure — through a draw from a Bernoulli distribution, whose parameter
θiin (t) is given by:
θiin (t) = 1 − e−ζ1 INi (t) ,
(4)
with 0 < ζ1 6 1. Note that according to 4, there are some scale-related returns to R&D
investment: access to innovative discoveries is more likely if a firm puts more resources into
in
R&D. If a firm innovates, it may draw a new machine embodying technology (Ain
i , Bi )
according to:
A
Ain
i (t) = Ai (t)(1 + xi (t))
Biin (t) = Bi (t)(1 + xB
i (t)),
B
where xA
i and xi are two independent draws from a Beta(α1 , β1 ) distribution over the
support [x1 , x1 ] with x1 belonging to the interval [−1, 0] and x1 to [0, 1]. Note that the
notional possibilities of technological advance — i.e. technological opportunities — are
captured by the support of the Beta distribution and by its shape. So, for example,
9
In the following, we assume all capital-producing firms to be identical in their R&D propensity. This
is not too far from reality: R&D intensities are largely sector specific and associated with the sector-wide
nature of innovative opportunities and modes of innovative search (more in Pavitt, 1984; Dosi, 1988;
Klevorick et al., 1995).
10
Firms on the technological frontier, lacking anyone to imitate, obviously invest all their R&D budget
in the search for innovations.
7
with low opportunities the largest probability density falls over “failed” innovations —
that is potential capital goods which are “worse” in terms of costs and performances
than those already produced by the searching firm. Conversely, under a condition of rich
opportunities, innovations which dominate incumbent technologies will be drawn with
high probability. As we shall show below, a crucial role of “Schumpeterian” technology
policies is precisely that of influencing opportunities and micro capabilities.
Alike innovation search, imitation follows a two steps procedure. The possibilities of
accessing imitation come from sampling a Bernoulli(θiim (t)):
θiim (t) = 1 − e−ζ2 IMi (t) ,
(5)
with 0 < ζ2 6 1. Firms accessing the second stage are able to copy the technology of one of
im
the competitors (Aim
i , Bi ). We assume that firms are more likely to imitate competitors
with similar technologies and we use a Euclidean metrics to compute the technological
distance between every pair of firms to weight imitation probabilities.
All firms which draw a potential innovation or imitation have to put it on production
or keep producing the incumbent generation of machines. Comparing the technology
competing for adoption, firms choose to manufacture the machine characterized by the
best tradeoff between price and efficiency. More specifically, knowing that consumptiongood firms invest following a payback period routine (see Section 2.3), capital-good firms
select the machine to produce according to the following rule:
min phi (t) + bch (Ahi , t) ,
h = τ, in, im,
(6)
where b is a positive payback period parameter (see Eq. 10 below). Once the type of
machine is chosen, we capture the imperfect information pervading the market assuming
that each firm sends a “brochure” with the price and the productivity of its offered
machines to both its historical (HCi ) clients and to a random sample of potential new
customers (N Ci ), whose size is proportional to HCi (i.e., N Ci (t) = γHCi (t), with 0 <
γ < 1).
2.3
The Consumption-Good Industry
Consumption-good firms produce a homogenous goods using capital (i.e. their stock of
machines) and labor under constant returns to scale. Firms plan their production (Qj )
according to adaptive demand expectations (Dje ):
Dje (t) = f (Dj (t − 1), Dj (t − 2), . . . , Dj (t − h)),
8
(7)
where Dj (t − 1) is the demand actually faced by firm j at time t − 1 (h positive integer)11 .
The desired level of production (Qdj ) depends on the expected demand as well as on the
desired inventories (Njd ) and the actual stock of inventories (Nj ):
Qdj (t) = Dje (t) + Njd (t) − Nj (t − 1),
(8)
with Njd (t) = ιDje (t), ι ∈ [0, 1]. The output of consumption-good firms is constrained by
their capital stock (Kj ). If the desired capital stock (Kjd ) — computed as a function of
the desired level of production — is higher than the current capital stock, firms invest
(EIjd ) in order to expand their production capacity12 :
EIjd (t) = Kjd (t) − Kj (t).
(9)
The capital stock of each firm is obviously composed of heterogenous vintages of
machines with different productivity. We define Ξj (t) as the set of all vintages of machinetools belonging to firm j at time t. Firms scrap machines following a payback period
routine. Through that, technical change and equipment prices influence the replacement
decisions of consumption-good firms13 . More specifically, firm j replaces machine Aτi ∈
Ξj (t) according to its technology obsolescence as well as the price of new machines:
RSj (t) =
Aτi
p∗ (t)
≤b ,
∈ Ξj (t) :
c(Ai,τ , t) − c∗ (t)
(10)
where p∗ and c∗ are the price of and unit cost of production upon the new machines.
Firms compute their replacement investment summing up the number of old machinetools satisfying Equation 1014 .
Consumption-good firms choose their capital-good supplier comparing the price and
productivity of the currently manufactured machine-tools they are aware of. As we mentioned above (cf. Section 2.2) the capital-good market is systematically characterized by
imperfect information. This implies that consumption-good firms compare “brochures”
describing the characteristics of machines only from a subset of equipment suppliers.
Firms then choose the machines with the lowest price and unit cost of production (i.e.,
pi (t) + bc(Aτi , t)) and send their orders to the correspondingly machine manufacturer.
11
For maximum simplicity, here we use the rule Dje (t) = Dj (t − 1). In Dosi et al. (2006) we check the
robustness of the simulation results employing more sophisticated expectation-formation rules. We found
that increasing the computational capabilities of firms does not significantly change either the average
growth rates or the stability of the economy. These properties still hold in the model presented here.
12
We assume that in any give period firm capital growth rates cannot exceed a fixed maximum threshold consistent with the maximum capital growth rates found in the empirical literature on firm investment
patterns (e.g. Doms and Dunne, 1998).
13
This in line with a large body of empirical analyses (e.g., Feldstein and Foot, 1971; Eisner, 1972;
Goolsbee, 1998) showing that replacement investment is typically not proportional to the capital stock.
14
Moreover, they also scrap the machines older than η periods (with η being a positive integer).
9
Machine production is a time-consuming process: capital-good firms deliver the ordered
machine-tools at the end of the period15 . Gross investment of each firm (Ij ) is the sum of
expansion and replacement investment. Pooling the investment of all consumption-good
firms one gets aggregate investment (I).
Consumption-good firms have to finance their investments as well as their production,
as they advance worker wages. In line with a growing number of theoretical and empirical
papers (e.g. Stiglitz and Weiss, 1992; Greenwald and Stiglitz, 1993a; Hubbard, 1998)
we assume imperfect capital markets. This implies that the financial structure of firms
matters (external funds are more expensive than internal ones) and firms may be credit
rationed. More specifically, consumption-good firms finance production using their stock
of liquid assets (N Wj ). If liquid assets do not fully cover production costs, firms borrow
the remaining part paying an interest rate r up to a maximum debt/sales ratio of Λ. Only
firms that are not production-rationed can try to fulfill their investment plans employing
their residual stock of liquid assets first and then their residual borrowing capacity16 .
Given their current stock of machines, consumption-good firms compute average productivity (πj ) and unit cost of production (cj ). Prices are set applying a variable markup
(µj ) on unit costs of production:
pj (t) = (1 + µj (t))cj (t).
(11)
Markup variations are regulated by the evolution of firm market shares (fj )17 :
fj (t − 1) − fj (t − 2)
µj (t) = µj (t − 1) 1 + υ
fj (t − 2)
,
with 0 6 υ 6 1.
The consumption-good market too is characterized by imperfect information (antecedents in the same spirits are Phelps and Winter, 1970; Klemperer, 1987; Farrel and
Shapiro, 1988; see also the empirical literature on consumers’ imperfect price knowledge
surveyed in Rotemberg, 2008). This implies that consumers do not instantaneously switch
to products made by more competitive firms. However, prices are clearly one of the key
determinants of firms’ competitiveness (Ej ). The other component is the level of unfilled
demand (lj ) inherited from the previous period:
Ej (t) = −ω1 pj (t) − ω2 lj (t),
15
(12)
Among the empirical literature investigating the presence of gestation-lag effects in firm investment
expenditures see e.g. Del Boca et al. (2008).
16
If investment plans cannot be fully realized, firms give priority to capital stock expansion, as compared to the substitution of old machines.
17
This is close to the spirit of “customer market” models originated by the seminal work of Phelps
and Winter (1970). See also Klemperer (1995) for a survey and the exploration of some important macro
implications by Greenwald and Stiglitz (2003).
10
where ω1,2 are positive parameter 18 . Weighting the competitiveness of each consumptiongood firms by its past market share (fj ), one can compute the average competitiveness of
the consumption-good sector:
E(t) =
F2
X
Ej (t)fj (t − 1).
j=1
Such variable represents also a moving selection criterion driving, other things being
equal, expansion, contraction and extinction within the population of firms. We parsimoniously model this market setup letting firm market shares evolve according to a “quasi”
replicator dynamics (for antecedents in the evolutionary camp cf. Silverberg et al., 1988;
Metcalfe, 1994a):
Ej (t) − E(t)
fj (t) = fj (t − 1) 1 + χ
E(t)
,
(13)
with χ > 019 .
The profits (Πj ) of each consumption-good firm reads:
Πj (t) = Sj (t) − cj (t)Qj (t) − rDebj (t),
where Sj (t) = pj (t)Dj (t) and Deb denotes the stock of debt. The investment choices of
each firm and its profits determine the evolution of its stock of liquid assets (N Wj ):
N Wj (t) = N Wj (t − 1) + Πj (t) − cIj (t),
where cIj is the amount of internal funds employed by firm j to finance investment.
2.4
Schumpeterian Exit and Entry Dynamics
At the end of each period we let firms with (quasi) zero market shares or negative net
assets die and we allow a new brood of firms to enter the markets. We keep the number
of firms fixed, hence any dead firm is replaced by a new one.
In line with the empirical literature on firm entry (Caves, 1998; Bartelsman et al.,
2005), we assume that entrants are on average smaller than incumbents, with the stock
of capital of new consumption-good firms and the stock of liquid assets of entrants in
18
Recall that consumption-good firms fix production according to their demand expectations, which
may differ from actual demand. If the firm produced too much, the inventories pile up, whereas if its
production is lower than demand plus inventories, its competitiveness is accordingly reduced.
19
Strictly speaking, a canonic replicator dynamics evolves on the unit simplex with all entities having
positive shares. Equation 13 allows shares to become virtually negative. In that case, the firm is declared
dead and market shares are accordingly re-calculated. This is what we mean by a “quasi-replicator”
dynamics. Note that an advantage of such formulation is that it determines at the same time changes in
market shares and extinction events.
11
both sectors being a fraction of the average stocks of the incumbents20 . Concerning the
technology of entrants, new consumption-good firms select amongst the newest vintages
of machines, according to the “brochure mechanism” described above. The process- and
product-related knowledge of new capital-good firms is drawn from a Beta distribution,
whose shape and support is shifted and “twisted’ according to whether entrants enjoy
an advantage or a disadvantage vis-à-vis incumbents21 . In fact, the distribution of opportunities fro entrants vs. incumbents is a crucial characteristics of different sectoral
technological regimes and plays a role somewhat akin to the distance from the technological frontier of entrants discussed in Aghion and Howitt (2007).
2.5
The Labor Market
The labor market is certainly not Walrasian: real-wage does not clear the market and
involuntary unemployment as well as labor rationing are the rules rather than the exceptions. The aggregate labor demand (LD ) is computed summing up the labor demand
of capital- and consumption-good firms. The aggregate supply (LS ) is exogenous and
inelastic. Hence aggregate employment (L) is the minimum between LD and LS .
The wage rate is determined by institutional and market factors, with both indexation
mechanisms upon consumption prices and average productivity, on the one hand, and,
adjustments to unemployment rates, on the others:
∆w(t)
=g
w(t − 1)
∆U (t)
∆AB(t)
∆cpi(t)
,
,
cpi(t − 1) AB(t − 1) U (t − 1)
,
(14)
where cpi is the consumer price index, AB is the average labor productivity, and U is the
unemployment rate22 .
2.6
Consumption, Taxes, and Public Expenditures
An otherwise black boxed public sector levies taxes on firm profits and worker wages or
on profits only and pays to unemployed workers a subsidy (wu ), that is a fraction of the
current market wage (i.e., wu (t) = ϕw(t), with ϕ ∈ (0, 1)). In fact, taxes and subsidies
20
The stock of capital of a new consumption-good firm is obtained multiplying the average stock of
capital of the incumbents by a random draw from a Uniform distribution with support [φ1 , φ2 ], 0 < φ1 , <
φ2 6 1. In the same manner, the stock of liquid assets of an entrant is computed multiplying the average
stock of liquid assets of the incumbents of the sector by a random variable distributed according to a
Uniform with support [φ3 , φ4 ], 0 < φ3 , < φ4 6 1.
21
More precisely, the technology of capital-good firms is obtained applying a coefficient extracted from
a Beta(α2 , β2 ) distribution to the endogenously evolving technology frontier (Amax (t), B max (t)), where
Amax (t) and B max (t) are the best technology available to incumbents.
∆w(t)
∆AB(t)
22
. Simulation results are robust to
For simplicity, we assume in the following that w(t−1)
= AB(t−1)
wage dynamics involving adjustment to inflation and unemployment. For more detailed modelizations
of the labor market in a evolutionary/ACE framework see e.g. Tesfatsion (2000); Fagiolo et al. (2004);
Neugart (2008).
12
are the fiscal leverages that contribute to the aggregate demand management regimes
(we shall explore this issue in more detail below). Note that a “zero tax, zero subsidy”
scenario is our benchmark for a pure Schumpeterian regime of institutional governance.
Aggregate consumption (C) is computed by summing up over the income of both
employed and unemployed workers:
C(t) = w(t)LD (t) + wu (LS − LD (t)).
(15)
The model satisfies the standard national account identities: the sum of value added of
capital- and consumption goods firms (Y ) equals their aggregate production since in our
simplified economy there are no intermediate goods, and that in turn coincides with the
sum of aggregate consumption, investment and change in inventories (∆N ):
F1
X
i=1
Qi (t) +
F2
X
Qj (t) = Y (t) ≡ C(t) + I(t) + ∆N (t).
j=1
The dynamics generated at the micro-level by decisions of a multiplicity of heterogenous, adaptive agents and by their interaction mechanisms is the explicit microfoundation
of the dynamics for all aggregate variables of interest (e.g. output, investment, employment, etc.). However, as the model amply demonstrates, the aggregate properties of
the economy do not bears any apparent isomorphism with those micro adjustment rules
outlined above. And a fundamental consequence is also that any “representative agent”
compression of micro heterogeneity is likely to offer a distorted account of both what
agents do and of the collective outcomes of their actions — indeed, well in tune with the
arguments of Kirman (1992) and Solow (2008).
3
Empirical Validation
The foregoing model does not allow for analytical, closed-form solutions. This general
ABM distinctive feature stems from the non-linearities present in agent decision rules
and their interaction patterns, and it forces us to run computer simulations to analyze
the properties of the stochastic processes governing the coevolution of micro and macro
variables23 . In what follows, we therefore perform extensive Monte-Carlo analyses to
wash away across-simulation variability. Consequently, all results below refer to acrossrun averages over one hundred replications and their standard-error bands24 .
23
Some methodological issues concerning the exploration of the properties of evolutionary/ACE models
are discussed in e.g. Lane (1993); Pyka and Fagiolo (2007); Fagiolo et al. (2007); Fagiolo and Roventini
(2008).
24
Preliminary exercises confirm that, for the majority of statistics under study, Monte-Carlo distributions are sufficiently symmetric and unimodal to justify the use of across-run averages as meaningful
synthetic indicators.
13
Let we start from a sort of “benchmark” setup for which the model is empirically
validated, i.e. it is studied in its ability to replicate a wide spectrum of microeconomic
and macroeconomic stylized facts. Initial conditions and parameters of the benchmark
setup are presented in Table 1.
As it should be clear from the forgoing presentation of the model, it embodies both a
Schumpeterian engine and a Keynesian one. The former rests in the generation of innovations by an ensemble of equipment producers which expensively search and endogenously
differentiate in the technology they are able to master. The Keynesian engine has two
parts: a direct one — through fiscal policies — and an indirect one — via investment decisions and workers’ consumption. Hence, the benchmark model appropriately embodies
all such Schumpeterian and Keynesian features.
Next we tune so to speak “up” and “down” the key policy variables (e.g. tax rates and
unemployment benefits) and we experiment with different conditions affecting the access
to and exploitation of new technological opportunities (e.g. the patent regime, anti-trust
policies).
Let us explore the ability of the model to reproduce the major stylized facts regarding
both the properties of macroeconomic aggregates and the underlying distribution of micro
characteristics (more on both in the direct antecedents to this model: [cf. Dosi et al., 2006,
2008).
Growth and Fluctuations. The model robustly generates endogenous self-sustained
growth patterns characterized by the presence of persistent fluctuations (cf. Figure 1). At
business cycle frequencies, bandpass-filtered output, investment and consumption series
(Bpf, cf. Baxter and King, 1999) display the familiar “roller-coaster” dynamics (see Figure
2) observed in real data (e.g. Stock and Watson, 1999; Napoletano et al., 2006). Moreover,
in tune with the empirical evidence, both consumption and investment appear to be
procyclical variables with the latter series being also more volatile than GDP.
Output, consumption and investment display strictly-positive average growth rates25
(cf. Table 2) and, according to Dickey-Fuller tests, they seem to exhibit a unit root. After
detrending the series with a bandpass filter, we compute standard deviations and crosscorrelations between output and the other series. In line with the empirical literature on
business cycles (cf. Stock and Watson, 1999), also in our model investment is more volatile
25
The average growth rate of variable X (e.g. GDP) is simply defined as:
GRX =
log X(T ) − log X(0)
,
T +1
where T = 600 is the econometric sample size. This value for T is a quite conservative choice, as the
first iterative moments of growth statistics converge to a stable behavior well before such a time horizon.
This means that the model reaches a relatively (meta) stable behavior quite soon after simulations start.
Our experiment show that choosing larger values for T does not alter the main economic implications of
the paper.
14
than output, whereas consumption is less volatile; consumption, investment, change in
inventories, and employment are procyclical; unemployment is countercyclical26 .
The model is also able to match the business-cycle properties concerning productivity,
price, inflation and markups: productivity is procyclical, prices are countercyclical and
leading; inflation is procyclical and lagging; markups are countercyclical (for the empirics
and discussion cf. Stock and Watson, 1999; Rotemberg and Woodford, 1999).
Finally, the aggregate growth rates of output display fat-tailed distributions well in
tune with the empirical evidence (cf. Table 4; see Castaldi and Dosi, 2008; Fagiolo et al.,
2008). Informally, that means that both in our model and in reality relatively big “spurs
of growth” and recessions occur much more frequently than it would be predicted on the
grounds of normally distributed shocks (see also below on firm growth patterns).
Distributions of Microeconomics Characteristics. Together with the ability of the
model to account for a rich ensemble of macro phenomena, how does it fare in its matching
with the evidence on the ubiquitous micro heterogeneity? Let us consider the regularities
concerning firm-size and growth-rate distributions, firm-productivity dynamics and firminvestment patterns which are generated by the model.
To begin with, well in tune with the empirical evidence (Dosi, 2007), rank-size plots
and normality tests suggest that cross-section firm (log) size distributions are skewed
and not log-normal (see Figures 3 and 4 and Table 5). Moreover, the estimation of the
shape parameters of exponential-power (Subbotin) distributions27 shows that pooled firm
growth-rate distributions are “tent-shaped” with tails fatter than the Gaussian benchmark
(see Table 4 and, for a comparison with the empirical evidence and some interpretation,
see Bottazzi and Secchi, 2003, 2006).
Turning to firm productivity, again in line with the empirical evidence (cf. the surveys
in Bartelsman and Doms, 2000; Dosi, 2007), firms strikingly differ in terms of labor
productivity (cf. standard deviations of labor productivities across firms plotted in Figure
5) and productivity differentials persist over time (cf. firm productivity autocorrelations
reported in Table 6)28 .
Finally, we have analyzed firm investment patterns. The model is indeed able to generate as an emergent property investment lumpiness (Doms and Dunne, 1998; Caballero,
1999). Indeed, in each time step, consumption-good firms with “near” zero investment
26
Consumption and net investment are also coincident variables matching yet another empirical regularity on business cycles. Changes in inventories are instead slightly lagging.
27
We estimate a distribution of the form:
f (x; b, a, m) =
1
1
b
2ab Γ(1 +
1 x−m b
a |
1
b)
e− b |
.
In a Subbotin distribution one parameter — b, in Table 4 — governs the fatness of the tails. The Normal
distribution is recovered when b = 2, the fatter Laplace distribution when b = 1.
28
In the last 200 periods of the simulations, we consider the autocorrelation of firms that survived for
at least 20 periods and we compute the industry average.
15
coexist with firms experiencing investment spikes (see Figure 6 and relate it to Gourio
and Kashyap, 2007).
4
Policy Experiments: Tuning Schumpeterian and
Keynesian Regimes
The model, we have seen, is empirically quite robust in that it accounts, together, for a
large number of empirical regularities. It certainly passes a much higher “testing hurdle”,
as Solow (2008) puts it, than simply reproducing “a few of the low moments of observed
time series: ratios of variances or correlation coefficients, for instance” (p. 245) as most
current models content themselves with. Encouraged by that empirical performance of
the model, let we experiment with different structural conditions (e.g. concerning the
nature of innovative opportunities) and policy regimes, and study their impact on output
growth rates, volatility and rates of unemployment29 .
4.1
Alternative Innovation and Competition Regimes
Consider first the Schumpeterian side of the economy, holding the “Keynesian engine”
constant as compared with the benchmark scenario30 : Table 7 summarizes the results.
Let us start by turning off endogenous technological opportunities. In this case, the
model collapses onto a barebone 2-sector Solow (1956) model in steady state, with fixed
coefficients and zero growth (absent demographic changes).
Opportunities and Search Capabilities. What happens if one changes the opportunities
of technological innovation and the ability to search for them? Experiment 1 (Table 7)
explores such a case. As compared to the benchmark, we shift rightward and leftward the
mass of the Beta distribution governing new technological draws (i.e. the parameters α1
and β1 , cf. Section 2.2). Note that the support of the distribution remains unchanged, so
that one could informally states that the notional possibilities of drift in the technological
frontier remain unchanged, too. However, the “pool” of opportunities agents actually face
get either richer or more rarefied. We find that higher opportunities have a positive impact
on the long-term rate of growth, reduce average unemployment and slightly increase GDP
volatility (a mark of Schumpeterian “gales of creative destruction”?).
Somewhat similarly, higher search capabilities approximated by the possibilities of accessing “innovations” — no matter if failed or successful ones — (cf. the ζ1,2 parameters in
29
Interestingly, most other statistical regularities concerning the structure of the economy (e.g. size
distributions, fatness of firms growth rates, etc.) appear to hold across an ample parameter range, under
positive technological progress, even when policies undergo the changes we study in the following.
30
The full list of parameters under different policy scenarios is available from the authors on request.
16
Equations 4 and 5) positively influence the rates of growth and lower unemployment. Together, business cycle fluctuations are dampened possibly because a population of “more
competent” firms entails lower degrees of technological asymmetries across them and indeed also lower degrees of “creative destruction”. See experiment 2, Table 7.
Note that such role of innovative opportunities and search capabilities is in principle equivalent to that blackboxed into the more aggregate notions of “human capital”
(Nelson and Phelps, 1966; Benhabib and Spiegel, 1994) and of “appropriate institutions”
(Acemoglu et al., 2006)31 .
Appropriability Conditions. In many current models with a (neo) Schumpeterian engine,
appropriability conditions play a key role via their assumptions on the forward looking rationality of the agent(s) investing into uncertain innovative search: the degrees of
monopoly appropriation of the economic benefits from successful search parametrize the
equilibrium relation between investment in R&D and rates of innovation. In this model,
we took a much more behavioural route and assumed a fixed propensity to invest in R&D
— again, quite in tune with the evidence displaying relatively sticky and sectoral specific
propensities. Granted that, how do changes in appropriability conditions affect aggregate
dynamics?
We first studied an extreme condition (albeit rather common in the theoretical literature), turning off the possibility of imitation, and assuming that all R&D is invested
in innovative search. Interestingly (and admittedly to the surprise of the authors) we
basically find no differences vis-à-vis the benchmark scenario: compare experiment 3.1
with experiment 0, Table 7.
Let we then try to mimic the effect of a patent system. Under a “length only” patent
scenario, the innovative technology cannot be imitated for a given number of periods
determined by the patent length (cf. experiment 3.2, Table 7). Such patenting possibility
is detrimental to long-run growth and also augments the average rate of unemployment.
The negative aggregate impact of the patent system is reinforced if each firm cannot
innovate in some neighborhood of the other firms’ technologies — i.e. in presence of a
patent breadth: see experiment 3.3, Table 7.
Entry and Competition Policies. Important dimensions of distinct Schumpeterian regimes
of innovation regard, first, the advantages/disadvantages that entrants face vis-à-vis incumbents and, second, the market conditions placing economic rewards and punishments
upon heterogenous competitors.
The first theme cuts across the evolutionary and neo-Schumpeterian literature and
sometimes is dramatized as a “Schumpeterian Mark I” vs. a “Schumpeterian Mark II”
scenarios, meaning systematic innovative advantages for entreprenurial entrants vs. cu31
In fact, given the increasing availability of micro data one can start thinking of disaggregates empirical proxies for our variables. The issue is however well beyond the scope of this work.
17
mulative advantages of incumbents (cf. Malerba and Orsenigo, 1995; Dosi et al., 1995).
In our model, technological entry barriers (or advantages) are captured by the probability distribution over the “technological draws” of entrants. Again, we hold constant the
support over which the economy (i.e. every firm thereof) may draw innovative advances,
conditional on the technology at any t. In this case we do it for sake of consistency:
results, even more so, apply if different regimes are also allowed to entail different probability supports. Let us first tune the Beta distribution parameters α2 and β2 (cf. Section
2.4). Our results are broadly in line with the evidence discussed in Aghion and Howitt
(2007): other things being equal, the easiness of entry and competence of entrants bears
a positive impact upon long-term growth, mitigates business cycles fluctuations and reduces average unemployment. See experiments 4.1 and 4.2, Table 7. However, the ceteris
paribus condition is equally important: the same aggregate growth patterns can be proved
to be equally guaranteed by competent cumulative learning of incumbents (see, above,
the exercises on search capabilities).
What about competitive conditions? An idea broadly shared across neoclassical and
evolutionary tradition is that “more competition is generally good”. In our model, that
translates somewhat narrowly into a less imperfect access to information to prices by
multiple heterogenous customers. In this case, in principle quite similar to the formally
slimmer Phelps and Winter (1970), higher competition is reflected by the higher replicator
dynamics parameter χ in Eq. 13. Simulation results suggest that a fiercer competition
has negligible effects on growth but it appears to reduce output volatility and average
unemployment (cf. exps. 5.1 and 5.2, Table 7).
Finally, we introduce antitrust policies by forbidding capital-good firms to exceed a
given market share (75% in experiment 6.1 and 50% in experiment 6.2, Table 7): the outcome is a lower unemployment rate, smaller business cycle fluctuations and also higher
GDP growth (on this point see also Fogel et al., 2008). Note that such a property have
little to do with any static “welfare gains” — which our model does not explicitely contemplates — but rather with the multiplicity of producers, and thus of innovative search
avenues, which antitrust policies safeguard32 .
4.2
The Keynesian Engine, or is Schumpeter Enough?
So far we have explored the effects of different means of “Schumpeterian” policies and
organizational setups over the long term (i.e. over the rate of growth of the economy) and
on unemployment rates and output volatilities. We did find significant effects both on the
long and short terms. However, to repeat, such results are conditional on a “Keynesian
machine” well in place. What happens if we switch that off? Remarkably, the system
dramatically slows down in terms of rates of growth, unemployment shoots up to utterly
32
The thrust of our results on policies affecting entry, competition, and variety preservation are indeed
broadly in tune with the advocacy for “evolutionary technology policies” in Metcalfe (1994b).
18
unrealistic levels and volatility increases. Note that it does so even if we consider a regime
with simultaneously “high opportunities” and “high search capabilities” (experiment 7,
Table 7). So for example, for the same richness of innovative opportunities, the long-term
rate of growth falls to around a third (compare experiments 7 and 1.2).
Let us further explore the role of fiscal policies over both the short- and long-term
properties of the economy. Consider the experiments presented in Table 8. We begin
with eschewing the public sector from the economy by setting both tax rate and unemployment benefits to zero while keeping the benchmark Schumpeterian characteristics in
place. In such a scenario, the economy experiments wilder fluctuations and higher unemployment rates in the short-run, but also an output growth in the long-run not far
from nil. Countercyclical Keynesian policies, as in the common wisdom, act indeed like
a parachute during recessions, sustaining consumption and, indirectly, investment on the
demand side. However, they also bear long-term effects on the supply side: in particular
on the rates of growth of productivity and output. Such a vicious feedback loop goes from
low output to low investment in R&D, low rates of innovation (cf. Equation 3) similar to
that pointed out by Stiglitz (1993)33 . In fact, in the latter as well as in our model, the
system may be “trapped” into a low growth trajectories which cannot be unlocked from
by a “Schumpeterian jumpstart”. Indeed, as we have seen above (experiment 7, Table 7)
Schumpeterian policies alone are not able to sustain high growth patterns and, even less
so, mild business cycle fluctuations and low unemployment.
Let us then allow for Keynesian demand macro-management policies and repeatedly
increase both the tax and the unemployment benefit rates. Tuning up fiscal demand management does delock the economy from the “bad” trajectory and brings it to the “good”
(high growth) one, which also our benchmark scenario happens to belong (cf. Table 8
and Figure 7). If one further increases the size of fiscal measures, average output growth
rates do not change as compared to the benchmark scenario, but output volatility and
unemployment significantly fall, and the economy spends more time in full employment
(cf. again Table 8 and Figure 7)34 .
5
Concluding Remarks
In this work we have studied the properties of an agent-based model that robustly reproduces a wide ensemble of macro stylized facts and distributions of micro characteristics.
The model entails the explicit account of search and investment decisions by populations of firms that are heterogeneous in the technologies which they master and, possibly,
in their decision rules. Aggregate macro properties are emergent from the thread of inter33
On the negative links between macroeconomic volatility, R&D investment, and long-run economic
growth in presence of financial market imperfections, see also Aghion et al. (2005, 2008).
34
On the long-run growth-enhancing effects of countercyclical macroeconomic policies, see the empirical evidence provided by Aghion and Marinescu (2007).
19
actions among economic agents, without any ex-ante consistency requirements amongst
their expectations and their actions. In that sense, the model may be considered an exercise in general disequilibrium analysis. Firms in the model endogenously generate new
technologies — embodied in new types of “machines” — via expensive and mistake-ridden
processes of search. Inventions then diffuse via the adoption decisions of machine users.
Hence, in the current statistical jargon, agents generate micro technological shocks and,
together, micro demand shocks which propagate through the economy.
In this respect, an important feature of the model is that it bridges Schumpeterian
theories of technology-driven economic growth with Keynesian theories of demand generation.
A central question that we address in the work is whether the “Schumpeterian engine”
by itself is able to maintain the economy on a high-growth / near full-employment path.
Broadly speaking, the answer is negative. Such an endogenous innovation engine is able
to do that only in presence of a “Keynesian” demand-generating engine, which in the
present model takes the form of public fiscal policies.
Our results also throw deep doubts on the traditional dichotomy between variables
impacting the long-run (typically, technology-related changes) and variables with a shortterm effect (traditional demand-related variables). On the contrary, technological innovations appear to exert their effects at all frequencies. Conversely, Keynesian demandmanagement policies do not only contribute to reduce output volatility and unemployment
rates, but for a large parameter region, they affect also long-run growth rates insofar as
they contribute to “delock” the economy from the stagnant growth trajectory which is
indeed one of the possible emergent meta-stable states.
The model appears to be a quite broad and flexible platform apt to perform a long
list of experiments, few of which have been presented above, studying the outcomes of
different policies and different institutional setups. An obvious direction of development
ought to address an explicit account of credit and financial markets (a somewhat germane
attempt in this direction is in Delli Gatti et al. (2005), broadly along Stiglitzian lines).
This is also a natural step in order to also analyze the real impact of monetary policies.
Another line of inquiry involves the comparison between alternative institutional specifications of the ways technologies are accessed and the ways markets work, somewhat along
the lines of the “variety of capitalism” approach (Soskice and Hall, 2001). More generally,
we view this as an example of a broader research program whereby explicit behavioral microfoundations nest the exploration of the relations between innovative dynamics, demand
generation, and policies affecting both.
References
Acemoglu, D., P. Aghion and F. Zilibotti (2006), “Distance to Frontier, Selection, and Economic Growth”,
Journal of the European Economic Association, 4: 37–74.
20
Aghion, P., M. Angeletos, A. Banerjee and K. Manova (2005), “Volatility and Growth: Credit Constraints
and Productivity-Enhancing Investment”, Working Paper 11349, NBER.
Aghion, P., P. Askenazy, N. Berman, G. Cette and L. Eymard (2008), “Credit Constraints and the
Cyclicality of R&D Investment: Evidence from France”, Working Paper 2008-26, Paris School of
Economics.
Aghion, P. and P. Howitt (1992), “A Model of Growth through Creative Destruction”, Econometrica, 60:
323–351.
Aghion, P. and P. Howitt (1998), Endogenous Growth, Cambridge, MIT Press.
Aghion, P. and P. Howitt (2007), “Appropriate Growth Policy: A Unifying Framework”, Journal of the
European Economic Association, 4: 269–314.
Aghion, P. and I. Marinescu (2007), “Cyclical Budgetary Policy and Economic Growth: What Do We
Learn from OECD Panel Data?”, in NBER Macroeconomics Annual 2007 volume 22 , 251–278.
Akerlof, G. A. (2002), “Behavioral Macroeconomics and Macroeconomic Behavior”, American Economic
Review, : 411–433.
Akerlof, G. A. (2007), “The Missing Motivation in Macroeconomics”, American Economic Review, 97:
5–36.
Akerlof, G. A. and J. L. Yellen (1985), “A Near-Rational Model of the Business Cycles, with Wage and
Price Inertia”, Quarterly Journal of Economics, 100: 823–838.
Bartelsman, E. and M. Doms (2000), “Understanding Productivity: Lessons from Longitudinal Microdata”, Journal of Economic Literature, 38: 569–94.
Bartelsman, E., S. Scarpetta and F. Schivardi (2005), “Comparative analysis of firm demographics and
survival: evidence from micro-level sources in OECD countries”, Industrial and Corporate Change, 14:
365–391.
Baxter, M. and R. King (1999), “Measuring Business Cycle: Approximate Band-pass Filter for Economic
Time Series”, The Review of Economics and Statistics, 81: 575–593.
Benhabib, J. and M. Spiegel (1994), “The Role of Human Capital in Economic Development: Evidence
from Aggregate Cross-Country Data”, Journal of Monetary Economics, 34: 143–173.
Blanchard, O. (2008), “The State of Macro”, Working Paper 14259, NBER.
Bottazzi, G. and A. Secchi (2003), “Common Properties and Sectoral Specificities in the Dynamics of
U.S. Manufacturing Firms”, Review of Industrial Organization, 23: 217–32.
Bottazzi, G. and A. Secchi (2006), “Explaining the Distribution of Firm Growth Rates”, RAND Journal
of Economics, 37: 235–256.
Caballero, R. J. (1999), “Aggregate Investment”, in J. Taylor and M. Woodford, (eds.), Handbook of
Macroeconomics, Amsterdam, Elsevier Science.
Canova, F. (2008), “How Much Structure in Empirical Models?”, in T. Mills and K. Patterson, (eds.),
Palgrave Handbook of Econometricsvolume 2, Applied Econometrics, Palgrave Macmillan.
Castaldi, C. and G. Dosi (2008), “The Patterns of Output Growth of Firms and Countries: Scale Invariances and Scale Specificities”, Empirical Economics, forthcoming.
Caves, R. (1998), “Industrial Organization and New findings on the Turnover and Mobility of Firms”,
Journal of Economic Literature, 36: 1947–1982.
21
Ciarli, T., A. Lorentz, M. Savona and M. Valente (2008), “Structural Change of Production and Consumption: A Micro to Macro Approach to Economic Growth and Income Distribution”, LEM Papers
Series 2008/08, Laboratory of Economics and Management (LEM), Sant’Anna School of Advanced
Studies, Pisa, Italy.
Colander, D. (ed.) (2006), Post Walrasian Macroeconomics, Cambridge, Cambridge University Press.
Colander, D., P. Howitt, A. P. Kirman, A. Leijonhufvud and P. Mehrling (2008), “Beyond DSGE Models:
Toward an Empirically Based Macroeconomics”, American Economic Review, 98: 236–240.
Dawid, H., S. Gemkow, P. Harting, K. Kabus, M. Neugart and K. Wersching (2008), “Skills, Innovation,
and Growth: An Agent-Based Policy Analysis”, Journal of Economics and Statistics, forthcoming.
Deissenberg, C., S. van der Hoog and H. Dawid (2008), “EURACE: A Massively Parallel Agent-based
Model of the European Economy”, Applied Mathematics and Computation, forthcoming.
Del Boca, A., M. Galeotti, C. P. Himmelberg and P. Rota (2008), “Investment and Time to Plan and
Build: A Comparison of Structures vs. Equipment in A Panel of Italian Firms”, Journal of the European
Economic Association, 6: 864–889.
Delli Gatti, D., C. Di Guilmi, E. Gaffeo, G. Giulioni, M. Gallegati and A. Palestrini (2005), “A New
Approach to Business Fluctuations: Heterogeneous Interacting Agents, Scaling Laws and Financial
Fragility”, Journal of Economic Behavior and Organization, 56: 489–512.
Dinopoulos, E. and P. Segerstrom (1999), “A Schumpeterian Model of Protection and Relative Wages”,
American Economic Review, 89: 450–472.
Doms, M. and T. Dunne (1998), “Capital Adjustment Patterns in Manufacturing Plants”, Review Economic Dynamics, 1: 409–29.
Dosi, G. (1988), “Sources, Procedures and Microeconomic Effects of Innovation”, Journal of Economic
Literature, 26: 126–71.
Dosi, G. (2007), “Statistical Regularities in the Evolution of Industries. A Guide through some Evidence
and Challenges for the Theory”, in F. Malerba and S. Brusoni, (eds.), Perspectives on Innovation,
Cambridge MA, Cambridge University Press.
Dosi, G., S. Fabiani, R. Aversi and M. Meacci (1994), “The Dynamics of International Differentiation: a
Multi-Country Evolutionary Model”, Industrial and Corporate Change, 3: 225–242.
Dosi, G., G. Fagiolo and A. Roventini (2006), “An Evolutionary Model of Endogenous Business Cycles”,
Computational Economics, 27: 3–34.
Dosi, G., G. Fagiolo and A. Roventini (2008), “The Microfoundations of Business Cycles: an Evolutionary,
Multi-Agent Model”, Journal of Evolutionary Economics, 18: 413–432.
Dosi, G., O. Marsili, L. Orsenigo and R. Salvatore (1995), “Learning, Market Selection and the Evolution
of Industrial Structures”, Small Business Economics, 7: 411–36.
Eisner, R. (1972), “Components of Capital Expenditures: Replacement and Modernization versus Expansion”, The Review of Economics and Statistics, 54: 297–305.
Fabiani, S., M. Druant, I. Hernando, C. Kwapil, B. Landau, C. Loupias, F. Martins, T. Mathä, R. Sabbatini, H. Stahl and A. Stokman (2006), “What Firms’ Surveys Tell Us about Price-Setting Behavior
in the Euro Area”, International Journal of Central Banking, 2: 3–47.
Fagiolo, G. and G. Dosi (2003), “Exploitation, Exploration and Innovation in a Model of Endogenous
Growth with Locally Interacting Agents”, Structural Change and Economic Dynamics, 14: 237–73.
Fagiolo, G., G. Dosi and R. Gabriele (2004), “Matching, Bargaining, and Wage Setting in an Evolutionary
Model of Labor Market and Output Dynamics”, Advances in Complex Systems, 14: 237–73.
22
Fagiolo, G., A. Moneta and P. Windrum (2007), “A Critical Guide to Empirical Validation of Agent-Based
Models in Economics: Methodologies, Procedures, and Open Problems”, Computational Economics,
30: 195–226.
Fagiolo, G., M. Napoletano and A. Roventini (2008), “Are Output Growth-Rate Distributions Fat-Tailed?
Some Evidence from OECD Countries”, Journal of Applied Econometrics, 23: 639–669.
Fagiolo, G. and A. Roventini (2008), “On the Scientific Status of Economic Policy: A Tale of Alternative
Paradigms”, Forthcoming in Knowledge Engineering Review.
Farrel, J. and C. Shapiro (1988), “Dynamic Competition with Switching Costs”, RAND Journal of
Economics, 19: 123–137.
Feldstein, M. and D. Foot (1971), “The Other Half of Gross Investment: Replacement and Modernization
Expenditures”, The Review of Economics and Statistics, 53: 49–58.
Fogel, K., R. Morck and B. Yeung (2008), “Corporate Stability and Economic Growth: Is What’s Good
for General Motors Good for America?”, Forthcoming in Journal of Financial Economics.
Fukac, M. and A. Pagan (2006), “Issues in Adopting DSGE Models for Use in the Policy Process”,
Working Paper 10/2006, CAMA.
Galı́, J. and M. Gertler (2007), “Macroeconomic Modelling for Monetary Policy Evaluation”, Journal of
Economic Perspectives, 21: 25–46.
Goolsbee, A. (1998), “The Business Cycle, Financial Performance, and the Retirement of Capital Goods”,
Working Paper 6392, NBER.
Gourio, F. and A. K. Kashyap (2007), “Investment Spikes: New Facts and a General Equilibrium Exploration”, Journal of Monetary Economics, 54: 1–22.
Greenwald, B. and J. Stiglitz (1993a), “Financial Market Imperfections and Business Cycles”, Quarterly
Journal of Economics, 108: 77–114.
Greenwald, B. and J. Stiglitz (1993b), “New and Old Keynesians”, Journal of Economic Perspectives, 7:
23–44.
Greenwald, B. and J. Stiglitz (2003), “Macroeconomic Fluctuations in an Economy of Phelps-Winter
Markets”, in P. Aghion, R. Frydman, J. Stiglitz and M. Woodford, (eds.), Knowledge, Information, and
Expectations in Modern Macroeconomics: In Honor of Edmund S. Phelps, Princeton, NJ, Princeton
University Press.
Hicks, J. R. (1937), “Mr. Keynes and the “Classics”: A Suggested Interpretation”, Econometrica, 5:
147–159.
Hubbard, G. R. (1998), “Capital-Market Imperfections and Investment”, Journal of Economic Literature,
36: 193–225.
Kirman, A. P. (1992), “Whom or What Does the Representative Individual Represent?”, Journal of
Economic Perspectives, 6: 117–136.
Klemperer, P. D. (1987), “Markets with Customer Switching Costs”, Quarterly Journal of Economics,
102: 375–394.
Klemperer, P. D. (1995), “Competition when Consumers have Switching Costs: An Overview with Applications to Industrial Organization, Macroeconomics and International Trade”, Review of Economic
Studies, 62: 515–539.
Klevorick, A. K., R. Levin, R. R. Nelson and S. G. Winter (1995), “On the Sources and Significance of
Interindustry Differences in Technological Opportunities”, Research Policy, 24: 185–205.
23
Lane, D. A. (1993), “Artificial Worlds and Economics, Part I and II”, Journal of Evolutionary Economics,
3: 89–107 and 177–197.
LeBaron, B. and L. Tesfatsion (2008), “Modeling Macroeconomies as Open-Ended Dynamic Systems of
Interacting Agents”, American Economic Review, 98: 246–250.
Louca, F. (2001), “Intriguing Pendula: Founding Metaphors in the Analysis of Economic Fluctuations”,
Cambridge Journal of Economics, 25: 25–55.
Malerba, F. and L. Orsenigo (1995), “Schumpeterian Patterns of Innovation”, Cambridge Journal of
Economics, 19: 47–65.
Metcalfe, J. S. (1994a), “Competition, Fisher’s Principle and Increasing Returns to Selection”, Journal
of Evolutionary Economics, 4.
Metcalfe, J. S. (1994b), “Evolutionary Economics and Technology Policy”, The Economic Journal, 104:
932–944.
Napoletano, M., A. Roventini and S. Sapio (2006), “Are Business Cycles All Alike? A Bandpass Filter
Analysis of the Italian and US Cycles”, Rivista Italiana degli Economisti, 1: 87–118.
Nelson, R. R. and E. S. Phelps (1966), “Investment in Humans, Technological Diffusion, and Economic
Growth”, American Economic Review, 61: 69–75.
Nelson, R. R. and S. G. Winter (1982), An Evolutionary Theory of Economic Change, Cambridge, The
Belknap Press of Harvard University Press.
Neugart, M. (2008), “Labor Market Policy Evaluation with ACE”, Journal of Economic Behavior and
Organization, , special Issue on “Agent-Based Models for Economic Policy Design”, edited by Dawid,
H. and Fagiolo, G., forthcoming.
Pavitt, K. (1984), “Sectoral Patterns of Technical Change: Towards a Taxonomy and a Theory”, Research
Policy, 13: 343–373.
Phelps, E. S. and S. G. Winter (1970), “Optimal Price Policy under Atomistic Competition”, in E. S.
Phelps, (ed.), Microeconomic Foundations of Employment and Inflation Theory, New York, Norton.
Pyka, A. and G. Fagiolo (2007), “Agent-Based Modelling: A Methodology for Neo-Schumpeterian Economics”, in H. Hanusch and A. Pyka, (eds.), The Elgar Companion to Neo-Schumpeterian Economics,
Cheltenham, Edward Elgar Publishers.
Romer, P. (1990), “Endogenous Technical Change”, Journal of Political Economy, 98: 71–102.
Rotemberg, J. (2008), “Behavioral Aspects of Price Setting, and Their Policy Implications”, Working
Paper 13754, NBER.
Rotemberg, J. and M. Woodford (1999), “The Cyclical Behavior of Prices and Costs”, in J. Taylor and
M. Woodford, (eds.), Handbook of Macroeconomics, Elsevier Science: Amsterdam.
Saviotti, P. and A. Pyka (2008), “Product Variety, Competition and Economic Growth”, Journal of
Evolutionary Economics, 18: 323–347.
Silverberg, G., G. Dosi and L. Orsenigo (1988), “Innovation, Diversity and Diffusion: A Self-Organization
Model”, The Economic Journal, 98: 1032–54.
Silverberg, G. and B. Verspagen (2005), “Evolutionary Theorizing on Economic Growth”, in K. Dopfer,
(ed.), Evolutionary Principles of Economics, Cambridge, Cambridge University Press.
Solow, R. M. (1956), “A Contribution to the Theory of Economic Growth”, Quarterly Journal of Economics, 70: 65–94.
Solow, R. M. (2008), “The State of Macroeconomics”, Journal of Economic Perspectives, 22: 243–246.
24
Soskice, D. and P. Hall (eds.) (2001), Varieties of Capitalism: the Institutional Foundations of Comparative Advantage, Oxford, Oxford University Press.
Stiglitz, J. (1993), “Endogenous Growth and Cycles”, Working Paper 4286, NBER.
Stiglitz, J. and A. Weiss (1992), “Credit Rationing in Markets with Imperfect Information”, American
Economic Review, 71: 393–410.
Stock, J. and M. Watson (1999), “Business Cycle Fluctuations in U.S. Macroeconomic Time Series”, in
J. Taylor and M. Woodford, (eds.), Handbook of Macroeconomics, Amsterdam, Elsevier Science.
Tesfatsion, L. (2000), “Structure, Behavior, and Market Power in an Evolutionary Labor Market with
Adaptive Search”, Journal of Economic Dynamics & Control, 25: 419–457.
Tesfatsion, L. and K. Judd (eds.) (2006), Handbook of Computational Economics II: Agent-Based Computational Economics, North Holland, Amsterdam.
Verspagen, B. (2002), “Evolutionary Macroeconomics: A Synthesis between Neo-Schumpeterian and
Post-Keynesian Lines of Thought”, The Electronic Journal of Evolutionary Modeling and Economic
Dynamics, 1007, http://www.e-jemed.org/1007/index.php.
Woodford, M. (2003), Interest and Prices: Foundations of a Theory of Monetary Policy, Princeton, NJ,
Princeton University Press.
25
Description
Number of firms in capital-good industry
Number of firms in consumption-good industry
R&D investment propensity
R&D allocation to innovative search
Firm search capabilities parameters
Beta distribution parameters (innovation process)
Beta distribution support (innovation process)
New-customer sample parameter
Capital-good firm mark-up rule
Desired inventories
Payback period
“Physical” scrapping age
Mark-up coefficient
Competitiveness weights
Replicator dynamics coefficient
Maximum debt/sales ratio
Interest Rate
Uniform distribution supports
(consumption-good entrant capital)
Uniform distribution supports
(entrant stock of liquid assets)
Beta distribution parameters
(capital-good entrants technology)
Tax rate
Unemployment subsidy rate
Symbol
F1
F2
ν
ξ
ζ1,2
(α1 , β1 )
[x1 , x1 ]
γ
µ1
ι
b
η
υ
ω1,2
χ
Λ
r
[φ1 , φ2 ]
Value
50
200
0.04
0.50
0.30
(3,3)
[−0.15, 0.15]
0.50
0.04
0.10
3
20
0.04
1
1
2
0.01
[0.10,0.90]
[φ3 , φ4 ]
[0.10,0.90]
(α2 , β2 )
(2, 4)
tr
ϕ
0.10
0.40
Table 1: Benchmark Parameters
Avg. growth rate
Dickey-Fuller test (logs)
Dickey-Fuller test (Bpf)
Std. Dev. (Bpf)
Rel. Std. Dev. (output)
Output
0.0254
(0.0002)
Consumption
0.0252
(0.0002)
Investment
0.0275
(0.0004)
6.7714
(0.0684)
−6.2564∗
(0.0409)
9.4807
(0.0957)
−5.8910∗
(0.0447)
0.2106
(0.0633)
−6.8640∗
(0.0905)
0.0809
(0.0007)
1
0.0679
(0.0005)
0.8389
0.4685
(0.0266)
5.7880
Table 2: Output, Investment, and Consumption Statistics. Bpf: bandpass-filtered
(6,32,12) series. Monte-Carlo simulation standard errors in parentheses. (∗): Significant
at 5%.
26
Series
(Bpf)
Output
Consumption
Investment
Net Investment
Ch. in Invent.
27
Employment
Unempl. Rate
Productivity
Price
Inflation
Mark-up
t-4
-0.1022
(0.0090)
-0.1206
(0.0123)
-0.2638
(0.0102)
-0.0838
(0.0122)
0.0072
(0.0081)
-0.3240
(0.0087)
0.3357
(0.0083)
0.1180
(0.0097)
0.2558
(0.0167)
-0.1070
(0.0151)
0.2183
(0.0118)
t-3
0.1769
(0.0080)
0.0980
(0.0129)
-0.3123
(0.0137)
0.0392
(0.0167)
0.1184
(0.0070)
-0.1901
(0.0123)
0.2084
(0.0118)
0.3084
(0.0088)
0.3181
(0.0218)
0.0841
(0.0135)
0.1599
(0.0088)
t-2
0.5478
(0.0048)
0.4256
(0.0106)
-0.2646
(0.0182)
0.2195
(0.0216)
0.2349
(0.0060)
0.0796
(0.0151)
-0.0596
(0.0147)
0.5316
(0.0092)
0.2702
(0.0235)
0.3110
(0.0175)
0.0411
(0.0128)
t-1
0.8704
(0.0014)
0.7563
(0.0062)
-0.0864
(0.0210)
0.4010
(0.0235)
0.2948
(0.0072)
0.4083
(0.0160)
-0.3923
(0.0158)
0.7108
(0.0093)
0.0916
(0.0216)
0.4456
(0.0226)
-0.0988
(0.0184)
Output (Bpf)
t
t+1
1
0.8704
(0)
(0.0014)
0.9527
0.9248
(0.0017) (0.0018)
0.1844
0.4473
(0.0206) (0.0175)
0.5114
0.5037
(0.0211) (0.0153)
0.2573
0.1331
(0.0090) (0.0098)
0.6692
0.7559
(0.0149) (0.0120)
-0.6607
-0.7550
(0.0148) (0.0120)
0.7672
0.6656
(0.0076) (0.0067)
-0.1645
-0.3950
(0.0198) (0.0212)
0.4021
0.1966
(0.0228) (0.0188)
-0.2040
-0.2361
(0.0213) (0.0206)
t+2
0.5478
(0.0048)
0.6848
(0.0038)
0.5950
(0.0139)
0.3850
(0.0103)
-0.0199
(0.0097)
0.6451
(0.0084)
-0.6489
(0.0084)
0.4378
(0.0097)
-0.5067
(0.0225)
-0.0628
(0.0154)
-0.1968
(0.0174)
t+3
0.1769
(0.0080)
0.3394
(0.0058)
0.5757
(0.0123)
0.2105
(0.0112)
-0.1319
(0.0085)
0.4067
(0.0069)
-0.4112
(0.0070)
0.1664
(0.0126)
-0.4688
(0.0210)
-0.2478
(0.0146)
-0.1226
(0.0135)
t+4
-0.1022
(0.0090)
0.0250
(0.0072)
0.4206
(0.0129)
0.0494
(0.0138)
-0.1640
(0.0067)
0.1555
(0.0082)
-0.1583
(0.0082)
-0.0609
(0.0128)
-0.3249
(0.0176)
-0.2900
(0.0131)
-0.0580
(0.0107)
Table 3: Correlation Structure. Bpf: bandpass- filtered (6,32,12) series. Monte-Carlo simulation standard errors in parentheses.
Series
Capital-good
Consumption-good
Output
b
0.5285
0.4249
1.4673
Std.Dev.
0.0024
0.0051
0.0122
a
0.4410
0.0289
0.0775
Std.Dev.
0.0189
0.0037
0.0004
m
-0.0089
0.0225
-0.0027
Std.Dev.
0.0002
0.0001
0.0003
Table 4: Growth-Rate Distributions, Estimation of Exponential-Power Parameters.
Industry
Capital-good
Consumption-good
Jarque-Bera
stat. p-value
20.7982
0
3129.7817
0
Lilliefors
stat. p-value
0.0464
0
0.0670
0
Anderson-Darling
stat. p-value
4.4282
0
191.0805
0
Table 5: Log-Size Distributions, Normality Tests
Industry
Capital-good
Consumption-good
t-1
0.5433
(0.1821)
0.5974
(0.2407)
t-2
0.3700
(0.2140)
0.3465
(0.2535)
Table 6: Average Autocorrelation of Productivity. Standard deviations in parentheses.
28
Exp.
0
Description
benchmark scenario
1.1
low technological opportunities
1.2
high technological opportunities
2.1
low search capabilities
2.2
high search capabilities
3.1
no imitation
3.2
patent (length only)
3.3
patent (breadth, too)
4.1
low entrant expected productivity
4.2
higher entrant expected productivity
5.1
weak market selection
5.2
strong market selection
6.1
weak antitrust
6.2
strong antitrust
7
Schumpeter-only, no fiscal policy
Avg. GDP
Growth Rate
0.0252
(0.0002)
0.0195
(0.0001)
0.0315
(0.0002)
0.0231
(0.0002)
0.0268
(0.0002)
0.0254
(0.0002)
0.0242
(0.0002)
0.0163
(0.0001)
0.0183
(0.0003)
0.0376
(0.0002)
0.0252
(0.0002)
0.0250
(0.0002)
0.0265
(0.0002)
0.0273
(0.0001)
0.0110
(0.0018)
GDP Std.
Dev. (Bpf)
0.0809
(0.0007)
0.0794
(0.0008)
0.0828
(0.0007)
0.0825
(0.0008)
0.0775
(0.0008)
0.0693
(0.0008)
0.0761
(0.0008)
0.0631
(0.0007)
0.0798
(0.0012)
0.0697
(0.0006)
0.0840
(0.0008)
0.0755
(0.0006)
0.0698
(0.0006)
0.0508
(0.0005)
1.5511
(0.0427)
Avg.
Unemployment
0.1072
(0.0050)
0.1357
(0.0050)
0.1025
(0.0051)
0.1176
(0.0059)
0.1031
(0.0048)
0.1049
(0.0059)
0.1132
(0.0060)
0.1329
(0.0067)
0.1402
(0.0084)
0.0853
(0.0047)
0.1253
(0.0058)
0.1024
(0.0053)
0.1036
(0.0043)
0.0837
(0.0036)
0.7855
(0.0274)
Table 7: Schumpeterian Regime Technological and Industrial Policy Experiments. Bpf:
bandpass-filtered (6,32,12) series. Monte-Carlo simulations standard errors in parentheses.
Tax Rate
0
Unemployment
Subsidy (in % of wages)
0
0.05
0.20
0.10
0.40
0.15
0.60
0.20
0.80
0.25
1
Avg. GDP
Growth Rate
0.0035
(0.0012)
0.0254
(0.0002)
0.0252
(0.0002)
0.0251
(0.0002)
0.0254
(0.0002)
0.0252
(0.0002)
GDP Std.
Dev. (Bpf)
1.5865
(0.0319)
0.1539
(0.0025)
0.0809
(0.0007)
0.0630
(0.0005)
0.0584
(0.0006)
0.0564
(0.0005)
Avg.
Unemployment
0.8868
(0.0201)
0.1952
(0.0086)
0.1072
(0.0050)
0.0846
(0.0034)
0.0602
(0.0027)
0.0551
(0.0023)
Table 8: Keynesian Regime Fiscal Policy Experiments. Bpf: bandpass- filtered (6,32,12)
series. Monte-Carlo simulations standard errors in parentheses.
29
25
Logs
20
15
10
5
GDP
Inv.
Cons.
0
0
50
100
150
200
250
Time
300
350
400
450
Figure 1: Level of Output, Investment, and Consumption (logs)
#
3
!'(
!')
!'&
.-/0-12
!'$
!
!!'$
!!'&
!!')
45.
617'
891:'
!!'(
!#3
!
"!
#!!
#"!
$!!
$"!
*+,-
%!!
%"!
&!!
&"!
Figure 2: Bandpass-Filtered Output, Investment, and Consumption
30
8
7
6
log(Rank)
5
4
3
2
Empirical
LogNormal
1
0
4
6
8
10
12
14
log(Size)
16
18
20
22
24
Figure 3: Pooled (Year-Standardized) Capital-good Firm Sales Distributions. Log Rank
vs. Log Size Plots.
10
9
8
7
log(Rank)
6
5
4
3
2
1
0
12
Empirical
LogNormal
14
16
18
log(Size)
20
22
24
Figure 4: Pooled (Year-Standardized) Consumption-good Firm Sales Distributions. Log
Rank vs. Log Size Plots.
31
15
Mean
Std. Dev.
Logs
10
5
0
0
50
100
150
Time
200
250
300
50
100
150
Time
200
250
300
15
Mean
Std. Dev.
Logs
10
5
0
0
Figure 5: Firms’ Productivity Moments (logs). First panel: capital-good firms. Second
panel: consumption-good firms.
1
I/K < 0.02
Percent
0.8
0.6
0.4
0.2
0
0
50
100
150
200
250
Time
300
350
400
450
1
I/K > 0.35
Percent
0.8
0.6
0.4
0.2
0
0
50
100
150
200
250
Time
300
350
400
450
Figure 6: Investment Lumpiness. First panel: share of firms with (near) zero investment;
second panel: share of firms with investment spikes.
32
percent
0.04
0.02
0
0
0.05
0.10
0.15
0.20
0.25
tax rate
0
0.05
0.10
0.15
0.20
0
0.05
0.10
0.15
0.20
0.25
tax rate
0
0.05
0.10
0.15
0.20
value
2
1
0
percent
1
unemp.
full. emp.
0.5
0
0
0.05
0.10
0.15
0.20
0.25
tax rate
0
0.05
0.10
0.15
0.20
Figure 7: Fiscal Policy Experiments. First panel: average output growth rate. Second
panel: bandpass-filtered output standard deviation. Third panel: average unemployment
rate (unemp.) and full-employment frequency (full emp.). In such policy experiments,
the unemployment subsidy rate (ϕ) is four times the tax rate.
33